> Are there fundamental reasons for not considering a decentralized
> design where each node has a representation of the graph?
I think it is important to differentiate what the "nodes" are that you
are referring to here.
If the nodes are social networking sites, then I would say yes. I
think the primary problem with this is a lack of control / privacy for
the user. If a Social Network has the ability to get all of your
data, you can't count on it to only access that which you would _like_
it to access.
If the nodes are the users, then I think that is the safest thing to
consider. The actual data shouldn't be decentralized, just the
users. For the user to truly own (and _control_) their data, they
should have a copy of their own connections/data and be given the
right to choose how and when to give it out. In order to do this,
they would need a tool that would allow them to aggregate their
content, then another tool (hopefully integrated with the first) that
allows them detailed control of how that data may be used by other
sources.
On Sep 22, 12:45 pm, "Steve G. Bjorg" <steve.bj...@gmail.com> wrote:
> On Sep 22, 2007, at 9:27 AM, Marc Danziger wrote:
> > We may "own" the social graph, but the reality of it is that it exists
> > on third-party servers who are, in essence aggregators of social data.
> > Those third parties may or may not have benign intentions, they may or
> > may not have legal protections against revealing social graphs to
> > government or private entities who seek them out, and they may or may
> > not have parallel interests to the individual 'owner' of the data. The
> > pressure to commercialize social network data will be intense (is
> > intense) and while I'm dipping into the middle of an ongoing
> > conversation about it, what I've not yet found in the discussion is a
> > core set of privacy principles that the social network services that
> > are proposed are architected around...pointers welcome, obviously.
> > On Sep 12, 11:08 am, "Gordon Rae" <gor...@premiumadvice.net> wrote:
> >>> I think it is important to make a key distinction about one
> >>> of the core goals of social network portability. My
> >>> understanding is that the objective is not to create a social
> >>> graph that is accessible to any party, but instead to empower
> >>> the user to have ownership of their social connections and be
> >>> able to take that information with them.
> >> That's very true, Max. let me see if I can scale it up a little.
> >> Our domain
> >> is a world of users, and we each have a social graph, which we
> >> own. We also
> >> have a variety of relationships with one another, with varying
> >> degrees of
> >> trust.
> > Marc
> Are there fundamental reasons for not considering adecentralized
> design where each node has a representation of the graph? In real
> life there isn't a concept of a graph holder. We each have our own
> piece of the graph and stitch it together. In such a design,
> everybody would own their part of the graph and federate it out (like
> trade?) for others to consume. An obvious downside will be
> consistency of the perceived global graph, but for large scale
> systems, this is impossible anyway.
> - Steve
> --------------
> Steve G. Bjorghttp://www.mindtouch.comhttp://www.opengarden.org
|
